MQL4 Random Entry Systems

[Jim]

It seems a number of people think we're crazy for publishing expert advisors that utilise random entries. Here we will demonstrate why we humbly disagree with that assertion.

We think that the way a lot of other people go about backtesting is crazy. Here we will also demonstrate how we think backtesting should be done, and the many pitfalls that lie in wait for the unwary. Highfield pointed out some of the limitations of the MetaTrader 4 strategy tester in this post, but we'll use it for now since a lot of people are familiar with it. If you're not familiar with it you can get hold of a copy free of charge and then learn how to use it without using up any of your precious seed corn capital.

We published our first example robot back in April 2009. We just dusted it off and ran it through the MetaTrader 4 strategy tester. We used an FXDD demo account so beloved of the Forex Megadroid team. That's because FXDD haven't yet got floating spreads and 5 decimal places like many other brokers these days, which makes it easier for other people to duplicate our results. Since cable has been doing well for us this year on our other systems we ran our backtest over the whole of 2009 on the GBP/USD currency pair. This is what the MetaTrader strategy tester showed us:

GuruEx01-2009-1.jpg

That's over 450% profit in a year. From the simplest "robot" we could devise this time last year, that stays in the market almost 100% of the time, and enters the market in the direction of the last tick.

Try it for yourself if you don't believe us:

[Mintaka]

Hi Jim,
I am probably one of those people who thought at some point in my Forex Experimental period, that backtesting is nonsense. Then I started to do it more profoundly and was better prepared (knowledge about the robots/ system) and wrote all the results down. After some time it started to make sense, and the randomness of the results become a kind of mathematical method. Even in the randomness you can find a kind of logic. And you can eliminate settings that are obviously wrong. However I didn’t use an expert adviser that utilise random entries. If something is random and provides random results you can expect a twofold randomness. But on the other hand most situations in our lives are random….maybe there is a method in that madness of yours?
But still ‘Was man nicht weiss, das eben brauchte man, und was man weiss, kan man nicht brauchen” (Goethe) in other words and in English “What you don’t know is far more relevant that what you do know”.

[Jim]

Hi Mintaka,

Thanks for your comment, and for your quotation! What do you make of mine from Einstein? Are you aware you could put those words of Goethe in your "signature" too? Maybe you know an even better quote you could put there instead!

To properly answer the points you raise will probably take several long threads! Sounds like you've studied some statistics, so I'll start by saying the entries aren't "random" in the sense of being generated using a random number generator. Two people backtesting the EA on historical data downloaded from the same MT4 broker ought to see the same sequence of trades. However because the EA simply enters based on the direction of the last tick, rather than at a particular price level or at a time determined by a combination of indicators, if you start your backtest running on a different day you will almost certainly see a completely different sequence of trades. Have you tried that? If so could you discern any "kind of logic" in the "twofold randomness of the results"?

One other thing to take away at the start is that evidently it is possible for such a "crazy" and "random" trading system to deliver profitable backtest results. Surely it must be possible to improve on its performance using indicators, filters, money management and all the other stuff you have probably heard about in countless forex courses and eBooks?

[Mintaka]

Thanks for your comment, and for your quotation! What do you make of mine from Einstein? Are you aware you could put those words of Goethe in your "signature" too? Maybe you know an even better quote you could put there instead!

Hi Jim, probably it sounds stupid, but how can I put a quote in my signature? Sorry from this frivolous question!

[Jim]

How can I put a quote in my signature?

Once you're logged in click where it says "Settings" near the top right of the screen. Then you should see a box entitled "My Settings" halfway down on the left hand side of the screen. The fifth line in that box should say "Edit Signature". Click that, then type in your chosen quotation. Finally click the button that says "Save Signature".

The only stupid questions are the ones you never ask!

[pipsondemand]

Hi Jim,

Many thanks for your kind and thoughtful response. I have to admit I never have much to do with experts as I explained, but it is a big worry when these things can be sold to unsuspecting traders. They must be making millions ! Their is actually a very serious side to this conversation. I've thought about it many times as to why there isn't more oversight of such abuses, I suppose the money back guarantee etc goes some way to protecting the consumer, but I should think many people buy these things and forget to make claims if the products didn't live up to expectation and the product sellers get away with it.

As you've proved to me today, they can look great very short term then WHAM lol Oh well, I'll continue to trade manually and sit for hours everyday in front of these bloody screens lol you can always live in hopes...

Many thanks,

Mark :-)

[Jim]

To provide some context. Mark originally asked me via email if I recommended trading one of our example expert advisors. Below is my reply. In brief, NO!!!!

All the experts available on our site are for educational purposes only. I wouldn't dream of using any of them for live trading!

Expert 2 stays in the market the whole time, and enters in the direction of the last tick. It's almost random! It uses a large stop loss, and it also uses Martingale position sizing. If you try enough backtests you'll discover that eventually it will blow up your account in the way you describe.

The point is this - If it's that easy to make a really dumb expert advisor that looks good in backtests, and even in live testing for a couple of months or more, what are all those "artificially intelligent robots" selling on ClickBank really worth?

[Jim]

Hi Mark,

I'm sure lots of people have bought "robots" over the internet and never bothered to ask for their money back. The vendors sell "the dream", then rely on the quirk in human behaviour that makes most people reluctant to admit their mistakes in order to "make their millions". To add insult to injury, the "manuals" that come with these things don't even begin to teach the unfortunate purchasers what's really involved in trying to make some money from trading.

Having said that, if you're already trading successfully manually don't give up hope just yet. It's certainly possible to automate some, or indeed all, of the processes you go through whilst sat in front of "those bloody screens" for hours on end. How much depends on how much of your experience can be broken down into a sequence of steps that can be programmed into a computer, and how much relies on "gut feel" or "intuition" or whatever you choose to call it. The sort of stuff that computers can't (as yet!) be programmed to do.

Jim

[Stan]

Hi Jim,
I am find the thread as very valuable for me. I have been testing the MT4 trading software for a several weeks
using your EA Example 05, but in short periods. MT4 ran just a few horus and made profit every time! It was very nice
for me and I started to think seriously about using the robot with a real Forex account.
Now I am sure, reading your posts in this thread, that it is very probable, I could loose a lot of money!
Thank you Jim for preserving me from loss!

I think, a demo account can be useful for two things, first to learn and familiarize with the trading platform
and with developing trading robots.
The second purpose is to test trading system with the historical data
and different values of input parameters.
However it makes no sense to watch running auto-trading software with a demo account
and real-time data to judge if the robot just at the moment makes profit or loss.
The just coming data, even in real-time, are in any case just historical, merely "fresh"

[Jim]

Hi Stan,

Thanks for your kind words.

I guess you must have got the example 5 source code from our blog, so for anyone else who might be interested I'm attaching it to this post also. It actually works in much the same way as robot example 1, posted at the start of this thread. Example 5 implements essentially the same (ludicrously simple!) trading system, but complicates matters by illustrating how to use MT4 pending orders.

You're correct about being in danger of losing money on a live account also. That's why all our examples say "It should NEVER be used on a live account" at the top! Apart from anything else, a robot let loose with even modest amounts of hard earned cash needs a whole lot more in the way of error detection and recovery than our examples.

That said, it's great to hear that the system has proved profitable for you in your tests! Would you mind sharing more details about exactly how you achieved those results?

Which currency pair(s) did you test, does your MetaTrader demo broker quote those pairs with 4 or 5 decimal places, and what numbers did you use for the input settings?

Cheers,

Jim